SC data entry: Packets 1-3

KW initial exploration (updated with data from 2018-01-31)

This is a document describing Kara’s initial exploration of the data from Packets 1-3 (!!!).

Setup

Look at differences by site

d_long_subscale_boot <- d_long_subscale %>%
  filter(!is.na(sum_score)) %>%
  group_by(ctry, packet, subscale) %>%
  do(data.frame(rbind(smean.cl.boot(.$sum_score)))) %>%
  ungroup() %>%
  filter(subscale != "attn") %>%
  left_join(d_long_subscale %>%
              filter(!is.na(sum_score)) %>%
              count(ctry, packet, subscale)) %>%
  mutate(packet = paste("packet", packet),
         ctry = factor(ctry,
                       levels = c("us", "ghana", 
                                  "thailand", "china", "vanuatu")))

Joining, by = c("ctry", "packet", "subscale")

ggplot(d_long_subscale_boot %>% 
         mutate(packet = gsub("packet ", "P", packet)),
       aes(x = ctry, y = Mean, color = ctry)) +
  facet_wrap(~ reorder(interaction(packet, subscale, sep = ": "), 
                       as.numeric(factor(packet))), 
             ncol = 5, scales = "free") +
  geom_pointrange(aes(ymin = Lower, ymax = Upper)) +
  geom_text(aes(label = paste0("(n=", n, ")"), y = Lower), 
            size = 2, nudge_x = 0.15, hjust = 0) +
  scale_x_discrete(expand = c(0, 1)) +
  scale_color_brewer(palette = "Dark2") +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90, hjust = 1),
        legend.position = "top") +
  labs(title = "mean scores by site",
       subtitle = "ordered by packet (P1, P2, P3)\nerror bars are bootstrapped 95% confidence intervals",
       x = "site", color = "site",
       y = "mean score (range varies by subscale)")

Look at correlations among subscales by site

d_long_subscale_boot %>%
  mutate(subscale = ifelse(grepl("invo_hardy_bentall", subscale),
                           paste(subscale,
                                 gsub("packet ", "p", packet),
                                 sep = "_"),
                           subscale)) %>%
  distinct(ctry, subscale, Mean) %>%
  ungroup() %>%
  spread(subscale, Mean) %>%
  data.frame() %>%
  column_to_rownames("ctry") %>%
  cor() %>%
  data.frame() %>%
  rownames_to_column("subscaleA") %>%
  gather(subscaleB, cor, -subscaleA) %>%
  ggplot(aes(x = subscaleA, y = subscaleB, fill = cor, label = round2(cor))) +
  geom_tile() +
  geom_text(size = 3) +
  scale_fill_distiller(guide = guide_colorbar(barheight = 6),
                       palette = "RdYlBu", limits = c(-1, 1)) +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
  labs(title = "correlations among mean subscale scores, by site",
       x = "",
       y = "")

Cluster analysis

cor_by_site <- d_long_subscale_boot %>%
  mutate(subscale = ifelse(grepl("invo_hardy_bentall", subscale),
                           paste(subscale,
                                 gsub("packet ", "p", packet),
                                 sep = "_"),
                           subscale)) %>%
  distinct(ctry, subscale, Mean) %>%
  filter(subscale != "invo_hardy_bentall_p3") %>%
  ungroup() %>%
  spread(subscale, Mean) %>%
  data.frame() %>%
  column_to_rownames("ctry")

… of subscales

clust_subscales <- hclust(dist(t(cor_by_site)))
ggdendrogram(clust_subscales) +
  labs(title = "hierarchical cluster of subscales",
       subtitle = "using mean subscale scores by site")

… of sites

clust_subscales <- hclust(dist(cor_by_site))
ggdendrogram(clust_subscales) +
  labs(title = "hierarchical cluster of sites",
       subtitle = "using mean subscale scores by site")

Look at correlations among subscales by individuals

Packet 1

d_long_subscale %>%
  filter(packet == 1, !is.na(sum_score)) %>%
  spread(subscale, sum_score) %>%
  select(-c(ctry, wher, recr, whoc, attn)) %>%
  mutate(subj = paste(subj, packet, version, sep = "_")) %>%
  select(-packet, -version) %>%
  distinct() %>%
  remove_rownames() %>%
  column_to_rownames("subj") %>%
  cor(use = "pairwise.complete.obs") %>%
  data.frame() %>%
  rownames_to_column("subscaleA") %>%
  gather(subscaleB, cor, -subscaleA) %>%
  ggplot(aes(x = subscaleA, y = subscaleB, fill = cor, label = round2(cor))) +
  geom_tile() +
  geom_text(size = 3) +
  scale_fill_distiller(guide = guide_colorbar(barheight = 6),
                       palette = "RdYlBu", limits = c(-1, 1)) +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
  labs(title = "PACKET 1: correlations among subscales",
       subtitle = "using pairwise complete observations",
       x = "",
       y = "")

Setting row names on a tibble is deprecated.

Packet 2

d_long_subscale %>%
  filter(packet == 2, !is.na(sum_score)) %>%
  spread(subscale, sum_score) %>%
  select(-c(ctry, wher, recr, whoc, attn)) %>%
  mutate(subj = paste(subj, packet, version, sep = "_")) %>%
  select(-packet, -version) %>%
  distinct() %>%
  remove_rownames() %>%
  column_to_rownames("subj") %>%
  cor(use = "pairwise.complete.obs") %>%
  data.frame() %>%
  rownames_to_column("subscaleA") %>%
  gather(subscaleB, cor, -subscaleA) %>%
  ggplot(aes(x = subscaleA, y = subscaleB, fill = cor, label = round2(cor))) +
  geom_tile() +
  geom_text(size = 3) +
  scale_fill_distiller(guide = guide_colorbar(barheight = 6),
                       palette = "RdYlBu", limits = c(-1, 1)) +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
  labs(title = "PACKET 2: correlations among subscales",
       subtitle = "using pairwise complete observations",
       x = "",
       y = "")

Setting row names on a tibble is deprecated.

Packet 3

d_long_subscale %>%
  filter(packet == 3, !is.na(sum_score)) %>%
  spread(subscale, sum_score) %>%
  select(-c(ctry, wher, recr, whoc, attn)) %>%
  mutate(subj = paste(subj, packet, version, sep = "_")) %>%
  select(-packet, -version) %>%
  distinct() %>%
  remove_rownames() %>%
  column_to_rownames("subj") %>%
  cor(use = "pairwise.complete.obs") %>%
  data.frame() %>%
  rownames_to_column("subscaleA") %>%
  gather(subscaleB, cor, -subscaleA) %>%
  ggplot(aes(x = subscaleA, y = subscaleB, fill = cor, label = round2(cor))) +
  geom_tile() +
  geom_text(size = 3) +
  scale_fill_distiller(guide = guide_colorbar(barheight = 6),
                       palette = "RdYlBu", limits = c(-1, 1)) +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
  labs(title = "PACKET 3: correlations among subscales",
       subtitle = "using pairwise complete observations",
       x = "",
       y = "")

Setting row names on a tibble is deprecated.

All packets

Note that some of these pairs of subscales probably have very few observations going into these correlations! Take this with a grain of salt.

d_long_subscale %>%
  spread(subscale, sum_score) %>%
  select(-c(ctry, wher, recr, whoc, attn)) %>%
  mutate(subj = paste(subj, packet, version, sep = "_")) %>%
  select(-packet, -version) %>%
  distinct() %>%
  remove_rownames() %>%
  column_to_rownames("subj") %>%
  cor(use = "pairwise.complete.obs") %>%
  data.frame() %>%
  rownames_to_column("subscaleA") %>%
  gather(subscaleB, cor, -subscaleA) %>%
  ggplot(aes(x = subscaleA, y = subscaleB, fill = cor, label = round2(cor))) +
  geom_tile() +
  geom_text(size = 3) +
  scale_fill_distiller(guide = guide_colorbar(barheight = 6),
                       palette = "RdYlBu", limits = c(-1, 1)) +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
  labs(title = "PACKETS 1-3: correlations among subscales",
       subtitle = "using pairwise complete observations",
       x = "",
       y = "")

Setting row names on a tibble is deprecated.